A multi-rule combination determination method for simulation training tasks

Abstract

The application relates to the technical field of demonstration tool systems, and discloses a multi-rule combination judgment method for simulation training tasks, which comprises the following steps: constructing an isolated judgment context for an independent judgment trigger by relying on a cluster agent, and transferring a standardized data structure by means of static one-way association of the cluster agent in a task layer; collecting a time variable and a space oscillation frequency of an activation state of a previous judgment trigger to determine a cumulative manipulation deviation factor by relying on a time sequence characteristic data arbitration unit; and inputting the factor into a subsequent judgment trigger to adjust a critical trigger parameter, and comparing judgment position characteristic data under an adjusted actual distance boundary. The application solves the problem that control loss cannot be precisely evaluated due to isolated multi-dimensional rules and state transmission conflicts, isolates rule judgment logic, realizes dynamic compensation of continuous operation, and improves the self-healing ability and operation stability of a judgment link.

Description

Technical Field

[0001] This invention belongs to the technical field of demonstration tool systems, and particularly relates to a multi-rule combination judgment method for simulation training tasks. Background Technology

[0002] Currently, in real-time sub-item evaluation of maneuvering behavior, aircraft and vehicle simulation training and evaluation systems generally adopt configurable triggers or state machine logic to collect high-frequency, multi-dimensional spatial maneuvering input data from operators. The input data is then compared with preset discrete thresholds within the system to output a pass or fail result for a single judgment node. However, in multi-subject collaborative training processes, due to the instability of the human control link, variations in operational details generated during the preceding attitude adjustment phase have a cumulative transmission effect on the operational window of subsequent tracking actions. To ensure the response speed of the evaluation system, mainstream designs typically forcibly divide the continuous multi-stage training process into independent processing blocks that do not interfere with each other in the time sequence. This static segmentation approach easily ignores the spatiotemporal evolution patterns between operational actions. When facing training subjects with long cycles and high continuous control accuracy requirements, the existing architecture inevitably leads to a technical contradiction between excessive consumption of system resources and reduced evaluation confidence in order to maintain the isolation of individual evaluations.

[0003] Such evaluation platforms are limited by the physical interaction of the hardware rollers, making it difficult to adapt to varied manipulation structures. At the software control level, existing evaluation and simulation methods also have shortcomings. For example, Chinese invention patent application CN119376375A discloses a full-scenario multi-level linkage simulation test system, method, electronic device, and storage medium, which constructs a virtual spatiotemporal benchmark to achieve spatiotemporal consistency synchronization of multi-dimensional information. However, existing technologies typically focus on overall system-level data fusion and physical interface conversion. When faced with non-ideal interactions or long-cycle dynamic evolution, the quantification of continuous control losses cannot capture the temporal lag and spatial oscillations caused by preceding unstable actions. Furthermore, it cannot feed back operational variations in situ to subsequent triggering logic to adaptively converge the judgment boundary, causing the confidence level of the processing link evaluation to decrease when operating conditions fluctuate drastically.

[0004] Therefore, how to construct a multi-rule linkage judgment mechanism that combines state integrity transparent transmission with temporal feature convergence arbitration, and precisely quantify continuous control loss and elastically converge trigger boundary under the architecture of rule and task isolation, has become the technical problem to be solved by this invention. Summary of the Invention

[0005] The present invention aims to solve the problem that the continuous control loss cannot be accurately evaluated due to the isolation of multidimensional rules and the conflict of state transmission.

[0006] In this technical solution, a multi-rule combination judgment method for simulation training tasks is used in a demonstration tool system, comprising:

[0007] Step S1: Based on the cluster agent, construct isolated decision contexts for multiple independent decision triggers, and use the globally unique identifier to statically and unidirectionally associate the cluster agent through the upper task layer to pass standardized data structures to each decision trigger, thereby constructing a fault isolation and fault tolerance control closed loop for the abnormal or timeout state of a single decision trigger.

[0008] Step S2: Relying on the time-series characteristic data arbitration unit inside the cluster agent, the duration variable and spatial axial oscillation frequency of the preceding judgment trigger in the active state are collected. The ratio of the duration variable to the baseline time consumption parameter is defined as the timeout deviation parameter. The timeout deviation parameter is multiplied by the spatial axial oscillation frequency and 1 is added to determine the dimensionless control index used to characterize the degree of control quality impairment of the trainees and defined as the cumulative manipulation deviation factor.

[0009] Step S3: Input the cumulative manipulation deviation factor as a feedback parameter into the configuration item of the subsequent judgment trigger, adjust the critical trigger parameter inside the subsequent judgment trigger, so that when the subsequent judgment trigger runs, it adjusts the actual trigger judgment distance boundary to the standard effective distance radius divided by the cumulative manipulation deviation factor, and performs position feature data comparison judgment and cascaded status output under the adjusted actual trigger judgment distance boundary.

[0010] Preferably, the fault isolation and fault tolerance control closed loop for a single decision trigger's abnormal or timeout state in step S1 includes: when the sub-agent corresponding to a specific decision trigger in the decision context does not return response data within the stored timeout threshold parameter, the cluster agent automatically allocates a preset fallback state, determines the corresponding decision trigger as untriggered and records the corresponding abnormal log, and at the same time maintains the state update and data aggregation flow of the other parallel running sub-agents in the cluster agent, so as to avoid the infinite waiting or state deadlock caused by the single point of decision logic failure spreading to the entire task scoring link.

[0011] Preferably, step S2 includes the following refined sub-steps: Step S21, collecting the duration variable of the preceding judgment trigger being continuously in the active state and the frequency of spatial axial oscillation generated when the trainee manipulates the simulation training system in the active state; Step S22, retrieving the benchmark time consumption parameter pre-stored in the current task scoring rules, calculating the ratio of the duration variable to the benchmark time consumption parameter to generate the timeout deviation parameter; Step S23, multiplying the timeout deviation parameter with the spatial axial oscillation frequency, and adding 1 to the product result to generate the cumulative manipulation deviation factor.

[0012] Preferably, the standardized data structure transmission in step S1 includes: receiving heterogeneous raw judgment data output from different rule modules in multiple independent judgment triggers, uniformly converting the heterogeneous raw judgment data into a normalized state data stream of a standardized data structure, the normalized state data stream containing a unified Boolean activation flag and a normalized score value, and returning the normalized state data stream to the task layer for performance scoring.

[0013] Preferably, the method further includes the following recording and control steps: Step S51, recording the historical sequence data of the cumulative manipulation deviation factor of the same trainee in multiple consecutive training cycles, and establishing an evolution data matrix corresponding to the training cycle; Step S52, calculating the rate of change of the cumulative manipulation deviation factor with the progression of the training cycle based on the historical sequence data, and obtaining a quantitative trend index used to characterize the growth trend of manipulation skill proficiency; Step S53, when the quantitative trend index is lower than the stored proficiency threshold, automatically lowering the benchmark value of the benchmark time consumption parameter stored in the subsequent training task, so as to realize the dynamic adaptive control of the simulation training difficulty.

[0014] Preferably, after outputting the cascaded status output in step S3, the cluster agent packages the output data of the generated cascaded status output and the cumulative manipulation deviation factor to generate a standardized incremental score package with a globally unique identifier, and transfers the standardized incremental score package to the management terminal to display the data source tracking map of each detailed score of the trainees on the management interface.

[0015] Preferably, in step S2, when collecting the duration variable and spatial axial oscillation frequency of the preceding determination trigger in the active state, the sampling period of the duration variable is locked within a closed value range of 10ms to 30ms, and the maximum count of the spatial axial oscillation frequency collected in a single operation within a single determination context period is limited to within 50 times, thereby ensuring the data convergence when calculating the cumulative manipulation deviation factor and the data stability of the underlying registers in the timing feature data arbitration unit.

[0016] Preferably, the aircraft simulation attitude system, which serves as the simulation object in the simulation training task, is connected to multiple independent decision triggers. The position feature data includes simulated three-dimensional coordinate data and simulated velocity vector data output by the attitude sensors in the aircraft simulation attitude system. The decision rules corresponding to the subsequent decision triggers include a logical combination of at least two of the following: hovering state decision rules, flight path deviation decision rules, and target lock decision rules.

[0017] Preferably, the output of the cascaded state in step S3 includes: the subsequent judgment trigger performs position comparison under the updated actual trigger judgment distance boundary, outputs a trigger activation signal when it is inside the actual trigger judgment distance boundary, outputs a trigger inactivation signal when it is outside the actual trigger judgment distance boundary, and performs AND logic gating operation on the trigger activation signal or trigger inactivation signal with the activation state signal of the preceding judgment trigger, and outputs the cascaded state judgment result representing the completion degree of the entire training link task.

[0018] Compared with existing technologies, the multi-rule combination judgment method for simulation training tasks of the present invention has the following advantages:

[0019] 1. In the multi-rule combination judgment of simulation training tasks, a unidirectional control link is constructed through the task layer, agent layer, and trigger layer. The trigger layer aggregates a single operation rule and sends four normalized states to the agent layer: inactive, active, triggered, and inactive. The agent layer transmits the normalized state to the task layer without loss through ordinary agents or cluster agents, so that the task layer is free from separate adaptation to heterogeneous rule output formats. This achieves decoupling and isolation between the rule judgment logic and the upper-level business subjects, so that the parameter adjustment of the operation rule and the design change of the subject process do not interfere with each other. While reducing the complexity of upper-level data parsing, it eliminates the risk of judgment link deadlock caused by the interweaving of multi-dimensional rules in the traditional architecture.

[0020] 2. Relying on the time-series characteristic data arbitration unit within the cluster agent, the duration variable and sampling anomaly frequency of the preceding trigger in the active state are collected. The cumulative manipulation deviation factor, which characterizes the degree of control quality loss of the trainee, is calculated through the nonlinear product of time overflow effect and spatial oscillation frequency. This factor is then written in reverse as a feedback instruction to the control port of the subsequent trigger to forcibly adjust its intrinsic critical judgment threshold. This causes the actual trigger judgment distance boundary of the subsequent trigger to follow linearly with the reciprocal of the factor, thereby implementing the comparison and judgment of position characteristic data within a reduced tolerance range to achieve a dynamic compensation mechanism for continuous operation.

[0021] 3. By running each trigger in an independent decision context, a fault isolation and fault tolerance control closed loop is constructed for the abnormal or timeout state of a single decision trigger. When a specific sub-agent does not return response data within the preset timeout threshold, the cluster agent automatically allocates a fallback state to determine it as an untriggered state and records the corresponding abnormal log. At the same time, the state update and data aggregation flow of the other parallel running sub-agents are maintained, avoiding the failure of a single point of decision logic from spreading to the entire task scoring link and causing the system to wait indefinitely or become deadlocked. This enhances the overall operational stability and fault self-healing capability of the multi-rule combination decision method without relying on idealized software operating conditions. Attached Figure Description

[0022] Figure 1 This is a flowchart of the simulation training multi-rule combination judgment process for the temporal feature arbitration of this invention;

[0023] Figure 2 This invention provides a data flow and state tracking diagram for the simulation training multi-rule decision architecture. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0025] A multi-rule combination judgment method for simulation training tasks, used in a demonstration tool system, includes:

[0026] Step S1: Based on the cluster agent, construct isolated decision contexts for multiple independent decision triggers, and use the globally unique identifier to statically and unidirectionally associate the cluster agent through the upper task layer to pass standardized data structures to each decision trigger, thereby constructing a fault isolation and fault tolerance control closed loop for the abnormal or timeout state of a single decision trigger.

[0027] Step S2: Relying on the time-series characteristic data arbitration unit inside the cluster agent, the duration variable and spatial axial oscillation frequency of the preceding judgment trigger in the active state are collected. The ratio of the duration variable to the baseline time consumption parameter is defined as the timeout deviation parameter. The timeout deviation parameter is multiplied by the spatial axial oscillation frequency and 1 is added to determine the dimensionless control index used to characterize the degree of control quality impairment of the trainees and defined as the cumulative manipulation deviation factor.

[0028] Step S3: Input the cumulative manipulation deviation factor as a feedback parameter into the configuration item of the subsequent judgment trigger, adjust the critical trigger parameter inside the subsequent judgment trigger, so that when the subsequent judgment trigger runs, it adjusts the actual trigger judgment distance boundary to the standard effective distance radius divided by the cumulative manipulation deviation factor, and performs position feature data comparison judgment and cascaded status output under the adjusted actual trigger judgment distance boundary.

[0029] Preferably, the fault isolation and fault tolerance control closed loop for a single decision trigger's abnormal or timeout state in step S1 includes: when the sub-agent corresponding to a specific decision trigger in the decision context does not return response data within the stored timeout threshold parameter, the cluster agent automatically allocates a preset fallback state, determines the corresponding decision trigger as untriggered and records the corresponding abnormal log, and at the same time maintains the state update and data aggregation flow of the other parallel running sub-agents in the cluster agent, so as to avoid the infinite waiting or state deadlock caused by the single point of decision logic failure spreading to the entire task scoring link.

[0030] Preferably, step S2 includes the following refined sub-steps: Step S21, collecting the duration variable of the preceding judgment trigger being continuously in the active state and the frequency of spatial axial oscillation generated when the trainee manipulates the simulation training system in the active state; Step S22, retrieving the benchmark time consumption parameter pre-stored in the current task scoring rules, calculating the ratio of the duration variable to the benchmark time consumption parameter to generate the timeout deviation parameter; Step S23, multiplying the timeout deviation parameter with the spatial axial oscillation frequency, and adding 1 to the product result to generate the cumulative manipulation deviation factor.

[0031] Preferably, the standardized data structure transmission in step S1 includes: receiving heterogeneous raw judgment data output from different rule modules in multiple independent judgment triggers, uniformly converting the heterogeneous raw judgment data into a normalized state data stream of a standardized data structure, the normalized state data stream containing a unified Boolean activation flag and a normalized score value, and returning the normalized state data stream to the task layer for performance scoring.

[0032] Preferably, the method further includes the following recording and control steps: Step S51, recording the historical sequence data of the cumulative manipulation deviation factor of the same trainee in multiple consecutive training cycles, and establishing an evolution data matrix corresponding to the training cycle; Step S52, calculating the rate of change of the cumulative manipulation deviation factor with the progression of the training cycle based on the historical sequence data, and obtaining a quantitative trend index used to characterize the growth trend of manipulation skill proficiency; Step S53, when the quantitative trend index is lower than the stored proficiency threshold, automatically lowering the benchmark value of the benchmark time consumption parameter stored in the subsequent training task, so as to realize the dynamic adaptive control of the simulation training difficulty.

[0033] Preferably, after outputting the cascaded status output in step S3, the cluster agent packages the output data of the generated cascaded status output and the cumulative manipulation deviation factor to generate a standardized incremental score package with a globally unique identifier, and transfers the standardized incremental score package to the management terminal to display the data source tracking map of each detailed score of the trainees on the management interface.

[0034] Preferably, in step S2, when collecting the duration variable and spatial axial oscillation frequency of the preceding determination trigger in the active state, the sampling period of the duration variable is locked within a closed value range of 10ms to 30ms, and the maximum count of the spatial axial oscillation frequency collected in a single operation within a single determination context period is limited to within 50 times, thereby ensuring the data convergence when calculating the cumulative manipulation deviation factor and the data stability of the underlying registers in the timing feature data arbitration unit.

[0035] Preferably, the aircraft simulation attitude system, which serves as the simulation object in the simulation training task, is connected to multiple independent decision triggers. The position feature data includes simulated three-dimensional coordinate data and simulated velocity vector data output by the attitude sensors in the aircraft simulation attitude system. The decision rules corresponding to the subsequent decision triggers include a logical combination of at least two of the following: hovering state decision rules, flight path deviation decision rules, and target lock decision rules.

[0036] Preferably, the output of the cascaded state in step S3 includes: the subsequent judgment trigger performs position comparison under the updated actual trigger judgment distance boundary, outputs a trigger activation signal when it is inside the actual trigger judgment distance boundary, outputs a trigger inactivation signal when it is outside the actual trigger judgment distance boundary, and performs AND logic gating operation on the trigger activation signal or trigger inactivation signal with the activation state signal of the preceding judgment trigger, and outputs the cascaded state judgment result representing the completion degree of the entire training link task.

[0037] Example 1: In a comprehensive mission assessment scenario for aircraft simulation training, when trainees perform target locking and effective strike missions, the system implements real-time judgment through a preset three-layer decoupled architecture. The trigger layer inside the system defines a locking trigger to represent the target locking state, an effective strike trigger to represent the strike effectiveness, and a survival trigger to represent the trainee's survival state. Each trigger runs in its own independent judgment context and uniformly outputs four types of normalized state data streams: inactive, active, triggered, or ineffective. The task layer statically associates the globally unique identifiers of successful and failed agents with the cluster agents in the agent layer, and the cluster agents perform multi-rule combination judgments.

[0038] When the trainee initiates the training task, the cluster agent concurrently reads the status of each sub-agent. While the locking trigger is in an active state, its internal timing characteristic data arbitration unit collects the duration variable of the trigger remaining in the active state in real time. This duration variable is then compared with the pre-stored standard baseline time consumption parameter in the task layer scoring rules to determine the timeout deviation parameter. This unit collects the frequency of spatial axial oscillations that cross the preset jitter threshold within the current preset sliding time window. By multiplying the timeout deviation parameter with this spatial axial oscillation frequency and adding 1, the floating-point arithmetic unit performs a binary floating-point multiplication instruction on the timeout deviation parameter (as a time term) and the spatial axial oscillation frequency (as a spatial counter), and accumulates the product by 1. This generates a value in a specified memory unit with a minimum value of 1.0 and a maximum value limited to 256.0 by the underlying register word length. The system uses a dimensionless floating-point characteristic damping parameter as the denominator control term in the downstream triggering logic. This parameter is used to calculate the standard judgment distance radius in the subsequent judgment rules in real time during program conditional branch jumps. The system execution flow then determines the cumulative manipulation deviation factor, which characterizes the degree of control quality impairment of the trainee. The logic of the product operation is to achieve dimensional normalization of manipulation loss by nonlinearly weighting the temporal redundancy through spatial instability. The timeout deviation parameter, as a dimensionless time ratio, characterizes the linear drift in the time dimension; the spatial axial oscillation frequency, as a counting feature, characterizes the kinetic energy disturbance in the spatial dimension. The product of the two constructs a comprehensive evaluation quantity reflecting the spatiotemporal coupling loss. The constant 1 is introduced to establish a logical benchmark, ensuring that the deviation factor is always equal to 1 under ideal manipulation conditions, so that the subsequent judgment boundary remains at the original value of the standard radius without deviation.

[0039] The cumulative manipulation deviation factor, as a dynamic feedback command, is written in-situ into the control port of the subsequent effective strike trigger. The standard effective distance radius inside the effective strike trigger, upon receiving the cumulative manipulation deviation factor, is divided by this factor, thereby dynamically compressing the actual trigger judgment distance boundary from the standard effective distance radius to an adjusted actual trigger judgment distance boundary. During dynamic compensation, adaptive contraction of the judgment tolerance is used to reduce control deviations caused by spatiotemporal oscillations in the preceding operation. Specifically, since the cumulative manipulation deviation factor generated during the preceding attitude adjustment stage reflects the loss of control accuracy, the subsequent judgment trigger, by narrowing the actual trigger judgment distance boundary, requires the trainee to perform the strike action within a higher-precision steady-state range to compensate for the decrease in overall system confidence caused by the instability of the preceding control. This contraction essentially uses the rigid increase of the judgment standard to lock the remaining control window, preventing the initial small deviation from being nonlinearly amplified in the cascaded link, thus achieving self-healing constraints and accuracy compensation for the impaired state of operation quality. The effective strike trigger, with its narrowed boundary after compression... Below, comparing the three-dimensional coordinate data and velocity vector data output in real time by the aircraft simulation attitude system, if the trainee's control actions meet the judgment conditions within the narrowed tolerance range, the trigger outputs a triggered state; if the boundary is extremely converged due to the oscillation of the preceding operation, and the trainee cannot converge the physical state to the boundary within the limited time, the trigger feedback is in a failed state. The cluster agent summarizes the cascaded output results of each sub-agent in real time, packages them with the cumulative control deviation factor, and generates a standardized incremental score package with a globally unique identifier, which is then transferred to the management terminal to achieve a precise quantitative evaluation of the implicit loss in the control process. After the cascaded state output, the cluster agent synchronously extracts the dynamic change of the actual trigger judgment distance boundary and sends a damping adjustment signal to the servo driver of the aircraft simulation attitude system. The driver increases the analog feedback torque of the joystick according to the magnitude ratio of the cumulative control deviation factor, presenting the physical resistance after the control accuracy is damaged in situ at the hardware interaction level, transforming the contraction of the numerical judgment boundary into a physical operation constraint that the trainee can perceive, and completing the closed-loop control from data calculation to hardware execution.

[0040] Example 2: This experiment objectively evaluates the response performance and accuracy of this technical solution under complex task conditions by verifying the multi-rule combination judgment of trainees in a simulated environment. The experimental platform relies on a general-purpose simulation computer, configured with a central processing unit with a main frequency of 3.2GHz, a memory capacity of 16GB, and an integrated simulated flight attitude system with a fixed output period of 10ms. The experimental and control groups are designed as follows: The experimental group of this invention is selected, which adopts the multi-rule combination judgment method. The parameters of each trigger are configured according to the benchmark values ​​set in the industry for primary flight subjects. The control group adopts hard-coded conventional judgment logic. In this logic, each rule is independent of the others, the judgment boundary is a fixed value, and it does not have the ability to dynamically respond to the operational quality of trainees. To simulate the operational fluctuations in real training, the experimental input signal is simulated in the simulation. Based on the three-dimensional coordinate data of the aircraft, random Gaussian noise with a mean of 0 and a standard deviation of 0.15 units was superimposed. During the experiment, when the trainees performed target locking and effective strike missions, the lock-on duration and spatial axial oscillation frequency were monitored. The extraction process of spatial axial oscillation frequency was as follows: the time-series feature data arbitration unit read the three-dimensional position vector output by the aircraft simulation attitude system in real time with a period of 10ms, and calculated the deviation vector between the current position vector and the preset standard track vector in Euclidean space; the first-order difference was performed on the components of the deviation vector on each coordinate axis, and the number of times the sign of the difference value was flipped and the flip amplitude exceeded the preset jitter threshold within the preset 300ms sliding window was counted. This number was defined as the spatial axial oscillation frequency, which was used to quantify the surface tremor frequency generated by the trainees when maintaining attitude balance.

[0041] Test data shows that, under the operating environment of the prototype of this invention, when the timing feature data arbitration unit detects continuous manipulation deviations in the trainee, the cumulative manipulation deviation factor slowly increases from the initial value of 1.05 to 1.42. The growth trend of this value is positively correlated with the axial instability of the manipulation command. The effective strike trigger dynamically compresses and adjusts the internal trigger judgment distance boundary from the standard 50 meters to 35.2 meters, that is, the actual trigger judgment distance boundary is equal to 50 meters divided by 1.42.

[0042] In the subsequent judgment process, the effective strike trigger of the present invention showed that its strike success judgment accuracy was 92.5% under the condition of high cumulative manipulation deviation factor. In contrast, the control group, under the same noise and manipulation fluctuation conditions, had a strike success judgment accuracy of only 78.1% due to the lack of a boundary dynamic adjustment mechanism. When the cumulative manipulation deviation factor climbed to above 2.5 due to extreme manipulation instability, the actual trigger judgment distance of the effective strike trigger was compressed to 20 meters. At this point, the output of the present invention was in a failed state, which could accurately eliminate false positives in the effective strike judgment caused by substandard operation quality. The data confirmed that the present technical solution effectively suppressed the impact of continuous manipulation loss on the stability of the judgment link by dynamically coupling operation quality and trigger boundary. Moreover, under the extreme condition of manipulation deviation factor exceeding 2.5, the system can achieve objective interception of false triggering behavior through boundary contraction, providing stable judgment support for simulation training tasks.

[0043] Example 3: In assessment scenarios for high-difficulty simulation training subjects, when trainees face sudden interference during target tracking and strike evaluation tasks, resulting in continuous deviations in their maneuvers, the system needs to address the risk of judgment standard drift caused by fluctuations in operational quality. In this scenario, the technical solution of this invention constructs a real-time judgment mechanism for continuous control loss to achieve precise quantification of operational behavior. During system implementation, the training task layer relies on a preset globally unique identifier to statically and unidirectionally associate a cluster agent to establish a judgment context. Lock triggers, effective strike triggers, and survival triggers serve as sub-judgment units, running in parallel within their respective independent contexts. The timing feature data arbitration unit within the cluster agent monitors the trainee's input signals for target manipulation in real time. When the lock trigger is in an active state, the timing feature data arbitration unit synchronously collects the duration variable of this active state and compares it with the pre-stored standard baseline time parameter in the task layer scoring rules to obtain the timeout deviation parameter. Simultaneously, within a 1500-millisecond sliding time window, the arbitration unit calculates the frequency of fluctuations in the maneuver command across a preset jitter threshold in the spatial axis.

[0044] The system multiplies the timeout deviation parameter with the spatial axial oscillation frequency and adds 1 to obtain the cumulative control deviation factor. This factor, as an internal state adjustment quantity, is written in-situ to the control port of the effective strike trigger. Upon receiving this factor, the effective strike trigger narrows and compresses the intrinsically set standard effective distance radius according to the reciprocal of the cumulative control deviation factor to obtain the actual trigger judgment distance boundary. For example, when the cumulative control deviation factor is calculated to be 1.25, the system reduces the initial standard effective distance radius of 50.0 meters to 40.0 meters through control logic. The effective strike trigger uses the three-dimensional coordinate data output in real time from the aircraft attitude system to compare the current strike distance with the narrowed actual trigger judgment distance boundary. If the trainee's control action makes the aircraft's coordinate data meet the judgment condition within the narrowed boundary range, the trigger determines that it has been triggered; otherwise, if the trainee's operation is unstable and the actual physical state cannot converge to the adjusted boundary, the trigger outputs a failed state.

[0045] Statistical data shows that under perturbation conditions with a manipulation deviation factor of 1.50, the sample group using this boundary dynamic convergence judgment logic achieved a hit success accuracy rate of 91.2%, with no misjudgments due to substandard operation quality. In contrast, the control group, under the same perturbation level, experienced a hit success accuracy rate of 75.4% due to the lack of a boundary dynamic adjustment mechanism. When the cumulative manipulation deviation factor climbed to 2.80 due to extreme operational instability, the actual trigger judgment distance was compressed to 17.8 meters, indicating that the trigger feedback of the sample group of this invention had failed. This logic ensures that the judgment link can adjust the evaluation scale in real time when manipulation loss occurs, realizing quantitative analysis of manipulation quality fluctuations in training tasks and self-healing control of the judgment logic.

[0046] Example 4: Before the simulation training task starts, the system executes a standardized baseline calibration procedure for various sensors and input terminals to ensure the stability of the judgment link under dynamic operating conditions. To ensure the compliance of the data processing flow, the system restricts the processing of the original control signals to the off-site operating environment of the simulator's local controller and collects the duration variable. frequency of spatial axial oscillation Before setting the basic parameters, the system calls a preset low-pass filtering algorithm to denoise the attitude sensor signal, extracting the effective manipulation components with a cutoff frequency between 5Hz and 15Hz. Simultaneously, the system replaces personal identification information with randomly generated task identifiers, ensuring that the decision chain only applies to neutral physical feature data streams. The operator inputs the standard calibration operation sequence into the simulation training system. The system calculates the inherent response delay of the device by comparing the input operation commands with the output simulation execution state. The system quantizes this response delay into a millisecond-level timing calibration factor and stores it in the verification parameter mapping table in local memory, serving as the benchmark for timing alignment of the input states of each decision trigger during subsequent task execution. If a decrease in the physical signal input quality of the training environment is detected during the training task execution, the system automatically triggers the data consistency verification process. By comparing the signals received by each decision trigger in the current sampling period with the preset standard behavioral characteristics, abnormal data fluctuation points are identified, and the fluctuation points are numerically reconstructed using normal sample data from adjacent periods.

[0047] To address the state interpretation conflicts arising from multi-user manipulation in multi-user collaborative modes, the system introduces a weighting mechanism based on the differences in the smoothness of trainees' manipulation history. The system pre-records the variance of each trainee's manipulation trajectory fluctuation within a preset training period. The reciprocal of this variance determines the trainee's manipulation stability evaluation weight. During the logic fusion phase of multi-rule combination judgment, when the system receives sets of operation instructions from different trainees, it performs weighted logical superposition on each instruction based on this evaluation weight. This ensures that the operation instructions of highly stable trainees have a higher logical contribution in the judgment chain. This weight coefficient is updated based on the trainee's manipulation stability evaluation results for that task period after each complete training task cycle, thereby achieving adaptive maintenance of the system's judgment logic and ensuring consistent training evaluation during long-term service. By establishing a mapping relationship between physical input parameters and evaluation weights, the system eliminates the influence of inconsistent hardware response characteristics and manipulation noise on multi-rule judgment results, establishing the operational benchmark for the simulation training task judgment logic.

[0048] Example 5: Before the simulation training task is deployed online, the system executes a standardized baseline calibration procedure for various sensors and input terminals to eliminate the impact of differences in physical response of different hardware devices on the judgment logic. The operator inputs the preset standard action command sequence into the simulation system. The system synchronously collects the physical displacement of the operating lever and the corresponding input port voltage signal, calculates the displacement conversion coefficient of each input channel through linear regression fitting, and stores it in memory as a hardware adaptation configuration file. During the execution of the training task, the system reads the original voltage input signal of the operating terminal in real time and converts it into standardized physical operation parameters in real time according to the displacement conversion coefficient.

[0049] When the system detects a transient jump in the joystick voltage signal that exceeds the preset specification range, it determines that the signal is an abnormal jitter caused by hardware contact resistance. It then triggers signal reconstruction logic, automatically extracting the average physical operation parameter from the previous stable sampling period to replace the abnormal value in the current period. This achieves trajectory smoothing of the control path. At the end of the task cycle, the system compares the historical control fluctuation variance of each trainee and calculates the control stability evaluation weight. For tasks involving multiple trainees, the system performs weighted logic superposition on multiple control input trajectories based on the control stability evaluation weight, establishing an effective control benchmark for task execution. The weight values ​​are adaptively and smoothly updated with the accumulation of training data in each task cycle, ensuring that the judgment criteria can respond to the evolution of the trainees' control stability. By establishing a mapping relationship between physical input parameters and evaluation weights, the system eliminates the influence of inconsistent hardware response characteristics and control behavior noise on the multi-rule judgment results, establishing the operating benchmark for the simulation training task judgment logic.

[0050] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A multi-rule combination judgment method for simulation training tasks, used in a demonstration tool system, characterized in that, include: Step S1: Based on the cluster agent, construct isolated decision contexts for multiple independent decision triggers, and use the globally unique identifier to statically and unidirectionally associate the cluster agent through the upper task layer to pass standardized data structures to each decision trigger, thereby constructing a fault isolation and fault tolerance control closed loop for the abnormal or timeout state of a single decision trigger. Step S2: Relying on the time-series characteristic data arbitration unit inside the cluster agent, the duration variable and spatial axial oscillation frequency of the preceding judgment trigger in the active state are collected. The ratio of the duration variable to the baseline time consumption parameter is defined as the timeout deviation parameter. The timeout deviation parameter is multiplied by the spatial axial oscillation frequency and 1 is added to determine the dimensionless control index used to characterize the degree of control quality impairment of the trainees and defined as the cumulative manipulation deviation factor. Step S3: Input the cumulative manipulation deviation factor as a feedback parameter into the configuration item of the subsequent judgment trigger, adjust the critical trigger parameter inside the subsequent judgment trigger, so that when the subsequent judgment trigger runs, it adjusts the actual trigger judgment distance boundary to the standard effective distance radius divided by the cumulative manipulation deviation factor, and performs position feature data comparison judgment and cascaded status output under the adjusted actual trigger judgment distance boundary.

2. The multi-rule combination judgment method for simulation training tasks according to claim 1, characterized in that, Step S1 involves constructing a fault isolation and fault tolerance control closed loop for a single decision trigger's abnormal or timeout state. This includes: when a sub-agent corresponding to a specific decision trigger within the decision context fails to return response data within the stored timeout threshold parameter, the cluster agent automatically adjusts the preset fallback state, determines the corresponding decision trigger as untriggered, and records the corresponding exception log. At the same time, it maintains the state update and data aggregation flow of the other parallel running sub-agents in the cluster agent to avoid the infinite waiting or state deadlock caused by the single point of failure of the decision logic spreading to the entire task scoring link.

3. The multi-rule combination judgment method for simulation training tasks according to claim 1, characterized in that, Step S2 includes the following detailed sub-steps: Step S21, collecting the duration variable of the pre-determining trigger being continuously in the active state and the frequency of spatial axial oscillation generated when the trainee manipulates the simulation training system in the active state; Step S22: Retrieve the baseline time consumption parameter pre-stored in the current task scoring rules, calculate the ratio of the duration variable to the baseline time consumption parameter to generate the timeout deviation parameter; Step S23: Multiply the timeout deviation parameter with the spatial axial oscillation frequency, and add 1 to the product result to generate the cumulative manipulation deviation factor.

4. The multi-rule combination judgment method for simulation training tasks according to claim 1, characterized in that, Step S1, which transmits the standardized data structure, includes: receiving heterogeneous raw decision data output from different rule modules in multiple independent decision triggers; uniformly converting the heterogeneous raw decision data into a normalized state data stream with a standardized data structure; the normalized state data stream contains a unified Boolean activation flag and a normalized score; and returning the normalized state data stream to the task layer for performance scoring.

5. The multi-rule combination judgment method for simulation training tasks according to claim 1, characterized in that, It also includes the following recording and control steps: Step S51, record the historical sequence data of the cumulative manipulation bias factor of the same trainee in multiple consecutive training cycles, and establish an evolution data matrix corresponding to the training cycle; Step S52: Calculate the rate of change of the cumulative manipulation deviation factor with the training cycle based on historical sequence data. Read the cumulative manipulation deviation factor values ​​corresponding to the most recent 5 consecutive training cycles from the evolution data matrix in the processor's data processing unit. Perform a first-order difference subtraction operation on the factor values ​​between adjacent training cycles to obtain 4 cycle increment components. Summate these 4 cycle increment components arithmetically and divide by 4 to obtain the average single-cycle variation step size. Take the negative of the average single-cycle variation step size as a discrete derivative characteristic value representing the rate of decrease in control loss and write it into the cache of the trend controller. This yields a quantitative trend index used to characterize the growth trend of manipulation skill proficiency. Step S53: When the quantitative trend index is lower than the stored proficiency threshold, automatically lower the baseline value of the baseline time parameter stored in subsequent training tasks to achieve dynamic adaptive adjustment of the simulation training difficulty.

6. The multi-rule combination judgment method for simulation training tasks according to claim 1, characterized in that, After outputting the cascaded status output in step S3, the cluster agent packages the output data of the generated cascaded status output and the cumulative manipulation deviation factor to generate a standardized incremental score package with a globally unique identifier, and transfers the standardized incremental score package to the management terminal to display the data source tracking map of each detailed score of the trainees on the management interface.

7. The multi-rule combination judgment method for simulation training tasks according to claim 1, characterized in that, In step S2, when collecting the duration variable and spatial axial oscillation frequency of the preceding decision trigger in the active state, the sampling period of the duration variable is locked within a closed value range of 10ms to 30ms, and the maximum count of the spatial axial oscillation frequency collected in a single operation within a single decision context period is limited to within 50 times. This ensures the data convergence when calculating the cumulative manipulation deviation factor and the data stability of the underlying registers in the timing feature data arbitration unit.

8. The multi-rule combination judgment method for simulation training tasks according to claim 1, characterized in that, The aircraft simulation attitude system, which serves as the simulation object in the simulation training task, is connected to multiple independent decision triggers. The position feature data includes simulated three-dimensional coordinate data and simulated velocity vector data output by the attitude sensors in the aircraft simulation attitude system. The decision rules corresponding to the subsequent decision triggers include a logical combination of at least two of the following: hovering state decision rules, flight path deviation decision rules, and target lock decision rules.

9. The multi-rule combination judgment method for simulation training tasks according to claim 1, characterized in that, Step S3 outputs the cascaded state output, which includes: the subsequent judgment trigger performs position comparison under the updated actual judgment distance boundary, outputs a trigger activation signal when it is inside the actual judgment distance boundary, and outputs a trigger deactivation signal when it is outside the actual judgment distance boundary. The trigger activation signal or trigger deactivation signal is then ANDed with the activation state signal of the preceding judgment trigger to perform a logic gating operation, and the cascaded state judgment result representing the completion degree of the entire training link task is output.

Images